Recently, Uber CEO Dara Khosrowshahi shared his insights on autonomous driving technology in a podcast hosted by Indian entrepreneur Nikhil Kamath. He believes that in the short term, Tesla's reliance solely on visual systems may struggle to achieve Superhero-level safety, while advocating for Waymo’s multimodal approach. Tesla CEO Elon Musk quickly responded on the X platform, emphasizing the benefits of a pure vision system, stating Camera FTW (Visual Lens for Victory).
Khosrowshahi emphasized during the interview that the safety performance of autonomous vehicles must exceed that of human drivers by multiple times to be deemed acceptable. He mentioned that Alphabet's Waymo has already proven that such safety standards are achievable. He stated: While Elon might think I’m wrong, I believe that creating a purely visual system capable of achieving superhero-level safety in the short term is a huge challenge. He also acknowledged that purely visual systems might emerge in the future but questioned why the cost-effective solid-state LiDAR technology, which could reduce expenses to $400 to $500 (approximately HKD 3,120 to HKD 3,900), isn't being considered to enhance safety. Khosrowshahi stressed that all of Uber's autonomous driving partners utilize a combination of visual cameras, radar, and LiDAR, and he believes this is the preferred solution.
In response to Khosrowshahi's perspective, Musk expressed his views on the X platform, reiterating Tesla's vision of a purely visual system. He mentioned: LiDAR sensors and traditional radar can clash, potentially compromising safety. If there's a discrepancy between the data from LiDAR and the visual cameras, which one should we rely on? Musk emphasized that the ambiguity of such sensors could increase risks, and pointed out: This is also why Waymo cannot safely operate on the roads. he revealed that Tesla has disabled the vehicle's radar system to enhance safety, concluding with Cameras for the win (Visual lenses triumph).
This debate highlights the fundamental divide in autonomous driving technology. Tesla's approach employs a pure vision system relying on eight cameras and a neural network, while Waymo is equipped with 29 cameras, six radar sensors, and five LiDAR units. Tesla argues that human drivers only depend on their eyes, making cameras and neural networks the sole pathway to achieving fully autonomous driving. In contrast, companies like Waymo believe that redundancy in sensors is essential for ensuring widespread safety in autonomous driving. In terms of cost, Tesla's camera system for each vehicle is approximately $400 (around HKD 3,120), while Waymo's fifth-generation sensor suite is estimated to cost as much as $12,700 (around HKD 99,060).
